Java Concurrency in Practice - Summary - Part 6

This is part 6 of  my notes from reading Java Concurrency in Practice.

NOTE: These summaries are NOT meant to replace the book.  I highly recommend buying your own copy of the book if you haven't already read it. 

Chapter 8 - Applying Thread Pools

1. In the Executor framework, there is an implicit coupling between tasks and execution policies.  Not all tasks are compatible with all execution policies.
2. If a task depends on the results of other tasks, then the execution policy must be carefully managed to avoid liveness problems.  Deadlocks can happen if the thread pool is bounded, i.e. thread starvation deadlock.
1. Will always deadlock if using Executors.newSingleThreadExecutor().
2. Other resources like JDBC connections may also be a bottleneck.
3. Document any pool sizing or configuration constraints.
3. Tasks that rely on thread confinement for thread-safety will not work well with thread pools.
4. Responsiveness of time-sensitive tasks may be bad if we use a single thread executor or if we submit several long running tasks to a small thread pool.  Use timed resource waits instead of unbounded waits.
5. Tasks that use ThreadLocal cannot be used with the standard Executor implementation as Executors may reuse or kill threads.  Do not use ThreadLocal to communicate value between tasks.
6. For compute-intensive tasks, an Ncpu-processor system achieves optimum utilization with a thread pool of Ncpu + 1 threads.  For tasks that include I/O or other blocking operations, use a larger thread pool since not all threads will be schedulable at all times.
7. ThreadPoolExecutor is the base class of executors returned by Executors.newCachedThreadPool, newFixedThreadPool and newScheduledThreadExecutor.  It is highly configurable.
8. We can specify the type of BlockingQueue that holds tasks awaiting execution.
1. unbounded LinkedBlockingQueue is the default for newFixedThreadPool and newSingleThreadExecutor.
2. Another option is to use a bounded LinkedBlockingQueue, ArrayBlockingQueue or PriorityBlockingQueue.
3. SynchronousQueue - not really a queue.  It is a mechanism for managing handoffs between threads.  Another thread must be waiting to accept handoff  - if pool maximum size has not been reached a new thread is created.  If no thread is available, the task is rejected.   Handoff is more efficient as we don't have to place the Runnable in an Queue. newCachedThreadPool uses a SynchronousQueue
9. newCachedThreadPool is a good default choice for an Executor.
10. Saturation Policy for a ThreadPoolExecutor can be modified by calling setRejectedExecutionHandler().
1. abort - causes execute() to throw the unchecked RejectedExecutionException.  Caller catches this exception and implements its own overflow handling.  This is the default.
2. discard - silently discard the newly submitted task.
3. discard-oldest - discard tasks that would be executed next and tries to resubmit the new task.
4. caller-runs - Tries to slow down the flow of new task submission by pushing some of the work to the caller.  It executes the newly submitted task not in a pool thread, but in the thread that calls execute().
5. There is no predefined saturation policy to make execute() block when the work queue is full.  However, this can be achieved using a Semaphore to bound the task injection rate. 
11. Thread Factories - whenever a thread pool needs to create a thread, it uses a thread factory.  ThreadFactory.newThread() is called whenever a thread pool needs to create a new thread.  Default thread factory creates a new non-daemon thread with no special configuration.  Use a custom thread factory to to specify an UncaughtExceptionHandler for pool threads, or instantiate an instance of a custom Thread class that does debug logging, or give pool threads more meaningful names.
12. Most ThreadPoolExecutor options can be changed after construction via setters.  Executors.unconfigurableExecutorService wraps an existing ExecutorService to ensure that its configuration cannot be changed further.  newSingleThreadExecutor() returns such a wrapped Executor rather than a raw ThreadPoolExecutor.  This is because newSingleThreadExecutor is implemented as a thread pool with one thread, and no one should be able to increase the pool size.
13. ThreadPoolExecutor was designed for extension.
1. beforeExecute and afterExecute hooks are called in the thread that executes the task.  Used for logging, timing, monitoring, statistics gathering.  Use ThreadLocal to share values between beforeExecute and afterExecute.
2. afterExecute is not called if task completes with an Error (regular exception is okay)
3. If beforeExcute throws a RuntimeException, the task is not executed and afterExecute() is not called.
4. terminated hook is called after the thread pool has shutdown - all tasks have finished and all worker threads have shut down.  Useful for releasing resources allocated by the Executor, notification, logging, finalize statistics gathering.
14. Parallelizing recursive algorithms
1. Sequential loops are suitable for parallelization when each iteration is independent of others, and the work done in each iteration is significant to offset cost of task creation.
2. Sequential loops within recursive algorithms can be parallelized.  Easier if iteration does not need value of recursive iterations it invokes.
3. In order to wait for all results, create a new Executor,  schedule the parallel tasks, call executor.shutdown() and then awaitTermination().

Java Concurrency in Practice - Summary - Part 5

This is part 5 of  my notes from reading Java Concurrency in Practice.

NOTE: These summaries are NOT meant to replace the book.  I highly recommend buying your own copy of the book if you haven't already read it. 

Chapter 7 - Cancellation & Shutdown

1. Java does not provide any mechanism for safely forcing a thread to stop what it is doing.  Instead, we need to rely on interruption where a thread can request another thread to stop.  The thread to be stopped may choose to ignore the request or can terminate after optionally performing cleanup operations.
2. Don't use the deprecated Thread.stop() and suspend() methods.
3. A task is cancelable if some external code can move it to a completed state before its normal completion. A task that supports cancellation must specify its cancellation policy:
1. how external code can request cancellation
2. responsiveness guarantees to cancellation requests
3. what happens on cancellation (such as cleanup operations)
4. One cooperative mechanism for terminating a task is to use a cancellation requested flag which the task periodically checks.  If the flag is set by some external code, then the task terminates early.  Remember to make the cancellation requested flag volatile.  Otherwise changes made by the external code may never become visible to the task.
1. The thread will exit only when it checks the cancellation flag.  Hence there is no guarantee on if and when the check will be made.
2. Can take a very long time to take effect (if at all) if the thread to be cancelled is stuck at a blocking operation.
5. Interruption Is usually the most sensible way to implement cancellation
6. Each thread has a boolean interrupted status, which is set to true when a thread is interrupted.
1. interrupt() interrupts the target thread
2. isInterrupted() returns the interrupted status of the target thread
3. interrupted() clears the interrupted status of the current thread and returns its previous value.  This is the only way to clear the interrupted status of a thread.  Poor choice of function name.
7. Blocking library calls try to detect when a thread has been interrupted and return early.  They clear the interrupted status and throw an InterruptedException.  There is no guarantee about how quickly a blocking method will detect interruption.  In practice, it happens fairly quickly.  When a thread that is not blocked is interrupted, its interrupted status is set.  It is upto the activity being cancelled to poll the interrupted status and respond appropriately. 
1.  A task does not need to immediately stop on detecting the interrupted status.  It can postpone acting on the interruption till a more opportune moment.  This can prevent internal data structures from being corrupted when interrupted in the middle of some critical operations.
8. There is a distinction between how tasks and threads react to interruption.  An interrupt on a worker thread in a thread pool can cancel the current task as well as shut down the worker thread. Hence, guesT code that doesn't own a thread must preserve the interrupted status of the thread after acting on the interrupt, so that the owner can appropriately deal with it later.
9. A thread should be interrupted only by its owner.  Because each thread has its own interruption policy, you should not interrupt a thread unless you know what interruption means to that thread.
10. Responding to interruption
1. Propagate the interruptedException, OR
2. Restore the interrupted status by calling Thread.currentThread.interrupt().  This is the only feasible solution in cases like Runnable.run() which does not allow exceptions to be thrown.
3. Only code that implements a thread's interruption policy may swallow an interruption request.  General purpose task and library code must never swallow interruption requests.
11. Activities that do not support cancellation but still call interruptible blocking methods must call them in a loop, retrying when interruption is detected.  The interruption status should be saved locally and restored before returning. Restoring the interrupted status immediately can result in an infinite loop.
12. Cancellation via Future
1. Future.cancel(boolean mayInterruptIfRunning)  - if mayInterruptIfRunning is true and the task is running in some thread, then that thread is interrupted.  If false, cancel() only means that don't run this task if it hasn't started yet.
2. Standard Executor implementation implement a thread interruption policy that allows tasks to be canceled through interruption.  Hence, it is ok to call Future.cancel(true) which interrupts the thread.
3. You should not interrupt a pool thread directly when attempting to cancel a task because you won't know what task is running at the time the interrupt is delivered. Cancel only through the task's Future.
4. When Future.get() throws an InterruptedException or TimeoutException and you know that the result is no longer required, cancel the task by calling Future.cancel().
13. Dealing with non-interruptile blocking
1. synchronous socket I/O in java.io - read/write in InputStream and OutputStream are not responsive to interruption, but closing the underlying socket makes any threads blocked in read/write to throw a SocketException.
2. Synchronous I/IO in java.nio - Interrupting a thread waiting on an InterruptibleChannel causes it to throw ClosedByInterruptionException and close the channel.  Closing an InterruptibleChannel cause threads blocked on channel operations to throw AsynchronousCloseException.
3. Asynchronous IO with Selector - A thread blocked in Selector.select() returns prematurely if close() or wakeup() is called.
4. Lock acquisition - A thread waiting for an intrinsic lock cannot be interrupted.  Explicit Lock class offers the lockInterruptibly method.
5. To perform non standard cancellation tasks (like closing a socket), override newTaskFor() in ThreadPoolExecutor to return a CancellableTask. CancellableTask extends Callable and overrides the FutureTask.cancel() method to close socket or perform any other nonstandard cancellation tasks.
14. Stopping a thread-based service
1. A thread pool owns the worker threads, and should be responsible for stopping them.  It should provide lifecycle methods that can be used by the application to shut down the pool, which in turn shuts down the worker threads.
15. ExecutorService provides shutdown() and shutdownNow(). shutdownNow() returns the list of tasks that had not started, so they can be logged or saved for future processing.  The returned Runnable objects may not be the same as what was submitted - they may be wrapped.
1. shutdownNow() provides no way of knowing the state of tasks in progress at shutdown time, unless the tasks themselves do checkpointing. Another option is to override execute() of AbstractExecutorService and pass in a wrapper Runnable that records tasks cancelled at shutdown.  There is a race condition that may cause a completed task to be marked as cancelled.  So tasks must be idempotent.
16. Poison Pill - a special object placed on the work queue is another way to convince a producer-consumer service to shut down.  Applicable only when the number of consumers and producers is known, and when the queue is unbounded.
17. Handling abnormal thread termination
1. Leading cause of premature thread death is RuntimeException.  If nothing special is done, the exception bubbles all the way up the stack and the thread is killed after printing a stacktrace to the console (which no one may be watching for)
2. If you are writing a worker thread class for a thread pool or executor service, be sure to catch Throwable and then notify the executor service of premature thread death.
3. Thread API provides an UncaughtExceptionHandler facility - when a thread exits due to an uncaught exception, the JVM reports this to an application provided UncaughtExceptionHandler.  Use Thread.setUncaughtExceptionHandler() to set the handler for the current thread or Thread.setDefaultUncaughtExceptionHandler to set it for all threads.
4. In long-running applications, always use uncaught exception handlers for all threads that at least log the exception.
5. To set an UncaughtExceptionHandler for pool threads, provide a ThreadFactory to the ThreadPoolExecutor.  Exceptions thrown from tasks make it to the UncaughtExceptionHandler only for tasks submitted with execute().  For those submitted with submit(), the exceptions are rethrown when calling Future.get()
18. JVM Shutdown
1. orderly shutdown - when the last nondaemon thread exits, System.exit(), Ctrl-C
2. abrupt shutdown - Runtime.halt(), SIGKILL the JVM
3. In orderly shutdown, JVM starts all shutdown hooks registered while Runtime.addShutdownHook(). Order of shutdown hook execution is not guaranteed.
4. After all shutdown hooks have completed, JVM may choose to run finalizers if runFinalizersOnExit is true,
5. JVM makes no attempt to stop or interrupt any application threads that are still running.
6. Shut-down hooks can run concurrently with other application threads.  So, they must be thread-safe.  Since JVM is shutting down, the application state may be messy. Hence the shut-down hooks must be coded extremely defensively.
7. Shut-down hooks should not use services that can be shutdown by the application or by other shutdown hooks.  One option is to use a single shutdown hook per application that executes various shutdown operations in sequence.
19. Daemon threads - existence of a Daemon thread does not prevent JVM from shutting down.  Internal JVM threads like GC thread are daemon threads.  When JVM exits, finally blocks of any existing daemon threads are not run.
1. Should be used sparingly, for activities that can be safely abandoned at any time without any cleanup.  
2. Do not use daemon threads for any tasks that perform I/O.
3. Generally used for housekeeping tasks like a background thread to remove expired cache entries.
20. Finalizers
1. GC treats objects that have a non-trivial finalize() method specially. finalize() is called after the memory is reclaimed.
2. Finalizers can run concurrently with application threads.  Hence, they must be thread-safe.
3. No guarantee about if or when they will run.
4. HIgh performance cost.
5. Usually, finally blocks and explicit close statements are sufficient to release resources, instead of using finalize().  
6. finalizers may be needed for objects that hold resources acquired by native methods.
7. Avoid finalizers

Java Concurrency in Practice - Summary - Part 4

This is part 4 of  my notes from reading Java Concurrency in Practice.

NOTE: These summaries are NOT meant to replace the book.  I highly recommend buying your own copy of the book if you haven't already read it. 

Chapter 6 - Task Execution

1. Individual client requests are a natural task boundary choice for server applications.
2. Creating a new thread per task is usually NOT a good idea.
1. Overheads of thread creation and teardown can add up to be quite significant.
2. Active threads consume system resources like memory even when they are idle, and also increase CPU contention.
3. Creating too many threads can result in an OutOfMemoryError.
3. Primary abstraction for task execution in Java is the Executor framework, not Threads.
1. Executor interface has a void execute(Runnable) methods.
4. Executors are the easiest way to implement a producer-consumer design.
5. Decouples task submission from execution.  Execution policy is separated from task submission.
6. Always use Executor instead of new Thread(runnable).start()
7. Different types of Thread Pools can be created using static factory methods of the Executors class:
1. newFixedThreadPool - creates new threads upto a maximum specified size.  Tries to keep pool size constant by adding new threads if some die due to  exceptions.
2. newCachedThreadPool - No bounds on number of threads.  Number of threads increase/decrease based on load.
3. newSingleThreadExecutor - Used to process tasks sequentially in order imposed by task queue (FIFO, LIFO, priority order).  Replaces thread if it dies unexpectedly.
4. newScheduledThreadPool - Fixed size thread pool that supports delayed and periodic task execution.
1. Use instead of Timer.
1. Timer creates only a single thread.  If one task takes too long, it affects the timing accuracy of subsequent TimerTasks.
2. An unchecked exception thrown by a TimerTask terminates the Timer thread.  The Timer thread is not resurrected, and the Timer is simply cancelled.
3.  Scheduled thread pools do not have the above two limitations  However,  Timers can schedule based on absolute time, while scheduled thread pools only support relative time.
8. Executor lifecycle has 3 states - running, shutting down, terminated.
9. ExecutorService interface (that extends Executor) offers methods like shutdown(), shutdownNow(), isShutdown(), isTerminated(), awaitTermination() to control Executor life cycle.
1. shutdown() - graceful shutdown.  Allow all running and previously submitted tasks to complete.  No new tasks are accepted.
2. shutdownNow() - cancel running tasks, and ignores any queued tasks.
3. Tasks submitted to executor after shutdown are passed to a rejection handler, which may silently drop the task or throw a RejectedExecutionException.
4. awaitTermination() is usually called immediately after calling shutdown().
10. ExecutorService.submit(Callable) returns a Future. A  Future represents the lifecycle of a task, and provides methods to monitor/control it.
11. CompletionService combines the functionality of an Executor and a BlockingQueue.  Submit a bunch of Callables to the Executor, and then wait for the results to be available using take() and poll().
1. An ExecutorCompletionService is a wrapper around an Executor - new ExecutorCompletionService(executor).
2. Tasks are submitted to the completion service, and not directly to the Executor.
3. Multiple completion services can share an Executor.
4. Keep track of the number of tasks submitted in order to determine the number of times to call take().
12. Future.get() supports a version that throws a TimeoutException if the result is not available within the specified time delay.
1. If a TimeoutException happens, then call cancel on the Future to cancel the task.  If the task is written to be cancelable, then it can be terminated to avoid consuming unnecessary resources.
13. ExecutorService.invokeAll - takes a collection of tasks and returns a collection of Futures.  Timed version of invokeAll returns when all tasks have completed, the calling thread is interrupted or if the timeout expires.  Use Future.isCancelled() to determine if a particular task completed or was interrupted/cancelled.